-
Notifications
You must be signed in to change notification settings - Fork 1
/
baseline.py
195 lines (149 loc) · 6.53 KB
/
baseline.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
from __future__ import print_function
import argparse
import torch
import torch.utils.data
from torch import nn, optim
from torch.nn import functional as F
from torchvision import datasets, transforms
from torchvision.utils import save_image
import matplotlib.pyplot as plt
import numpy as np
import scipy.stats
import math
from util import *
parser = argparse.ArgumentParser(description='DeepCAMA MNIST Example')
parser.add_argument('--train', type=str2bool, required=True, metavar='T/F')
parser.add_argument('--batch-size', type=int, default=128, metavar='N',
help='input batch size for training (default: 128)')
parser.add_argument('--epochs', type=int, default=100, metavar='N',
help='number of epochs to train (default: 10)')
parser.add_argument('--no-cuda', action='store_true', default=False,
help='disables CUDA training')
parser.add_argument('--seed', type=int, default=1, metavar='S',
help='random seed (default: 1)')
parser.add_argument('--log-interval', type=int, default=10, metavar='N',
help='how many batches to wait before logging training status')
args = parser.parse_args()
args.cuda = not args.no_cuda and torch.cuda.is_available()
torch.manual_seed(args.seed)
device = torch.device("cuda" if args.cuda else "cpu")
kwargs = {'num_workers': 1, 'pin_memory': True} if args.cuda else {}
train_data = datasets.MNIST('./data/train/', train=True, download=True, transform=transforms.ToTensor())
train_data_clean = list(train_data)
for idx, _ in enumerate(train_data_clean):
L1 = list(train_data_clean[idx])
L1.append(0)
train_data_clean[idx] = tuple(L1)
#train_loader = torch.utils.data.DataLoader(train_data_clean, batch_size=args.batch_size, shuffle=True, **kwargs)
vertical_shift_range = np.arange(start=0.0,stop=1.0,step=0.1)
listofones = [1] * 10
split_list = [int(element * 0.1*len(train_data)) for element in listofones]
train_data_aug_split = torch.utils.data.random_split(train_data, split_list, generator=torch.Generator().manual_seed(42))
train_data_aug = []
for idx, _ in enumerate(train_data_aug_split):
split_part = list(train_data_aug_split[idx])
for idx2, _ in enumerate(split_part):
#print(split_part[idx2][0].size())
shift_fn = transforms.RandomAffine(degrees=0,translate=(0.0,vertical_shift_range[idx]))
L1 = list(split_part[idx2])
L1[0] = shift_fn(L1[0])
L1.append(1)
split_part[idx2] = tuple(L1)
#train_data_aug_split[idx][idx2] = tuple(L1)
#print(L1)
train_data_aug = train_data_aug + split_part
#for idx, split_part in enumerate(train_data_aug_split):
# #print(list(split_part)[0][0].size())
# train_data_aug = train_data_aug + list(split_part)
#for idx, _ in enumerate(train_data_aug_split):
# #print(list(split_part)[0][0].size())
# split_part = list(train_data_aug_split[idx])
# train_data_aug = train_data_aug + split_part
train_data_clean_aug = train_data_clean + train_data_aug
train_loader = torch.utils.data.DataLoader(train_data_clean_aug, batch_size=args.batch_size, shuffle=True, **kwargs)
print(len(train_data_aug))
print(len(train_data_clean_aug))
test_data = datasets.MNIST('./data/train/', train=False, transform=transforms.ToTensor())
test_loader = torch.utils.data.DataLoader(test_data, batch_size=args.batch_size, shuffle=True, **kwargs)
# ------------------------------------------------------\
class DeepCAMABaseline(nn.Module):
def __init__(self):
super(DeepCAMABaseline, self).__init__()
self.fc1 = nn.Linear(28*28,512)
self.fc2 = nn.Linear(512,256)
self.fc3 = nn.Linear(256,126)
self.fc4 = nn.Linear(126,512)
self.fc5 = nn.Linear(512, 10)
self.d1 = nn.Dropout(p=0.25)
self.d2 = nn.Dropout(p=0.5)
def forward(self, x):
a1 = self.d1(F.relu(self.fc1(x.view(-1,28*28))))
a2 = self.d1(F.relu(self.fc2(a1)))
a3 = self.d1(F.relu(self.fc3(a2)))
a4 = self.d2(F.relu(self.fc4(a3)))
a5 = self.fc5(a4)
return a5
def train(epoch):
train_loss = 0
for batch_id, (data,y,clean) in enumerate(train_loader):
data = data.to(device)
y = y.to(device)
model.train()
optimizer.zero_grad()
x_recon = model(data)
loss = loss_function(x_recon,y)
loss.backward()
train_loss += loss.item()
optimizer.step()
if batch_id % args.log_interval == 0:
print('Train Epoch: {} [{}/{} ({:.0f}%)]\tLoss: {:.6f}'.format(
epoch, batch_id * len(data), len(train_loader.dataset),
100. * batch_id / len(train_loader),
loss.item() / len(data)))
print('====> Epoch: {} Average loss: {:.4f}'.format(
epoch, train_loss / len(train_loader.dataset)))
return
def pred(x):
out = model(x.to(device))
#y = y.detach().cpu().numpy()
out = out.detach().cpu().numpy()
label = np.argmax(out,axis = 1)
return label
model = DeepCAMABaseline().to(device)
optimizer = optim.Adam(model.parameters(), lr=(1e-4+1e-5)/2)
loss_function = nn.CrossEntropyLoss()
if __name__ == "__main__":
vertical_shift_range = np.arange(start=0.0,stop=1.0,step=0.1)
accuracy_list = [0]*vertical_shift_range.shape[0]
if args.train:
for epoch in range(1, args.epochs + 1):
train(epoch)
torch.save(model.state_dict(), 'baseline_aug.pt')
else:
index = 0
model.load_state_dict(torch.load('baseline_aug.pt', map_location=device))
model.eval()
for vsr in vertical_shift_range:
temp = 0
total_i = 0
#if (vsr <= 0.11 and vsr >= 0.09):
#print('here')
for i, (data, y) in enumerate(test_loader):
#if (data.size()[0] == args.batch_size): #resolve last batch issue later.
data, y = shift_image(x=data,y=y,width_shift_val=0.0,height_shift_val=vsr)
y_pred = pred(data)
#print(y,y_pred)
#print(y_pred)
y_temp = y.detach().cpu().numpy()
aa = accuracy(y_temp,y_pred)
temp = temp + aa
total_i = total_i + 1
#print(aa)
print(temp/total_i)
accuracy_list[index] = temp/total_i
index = index + 1
print(temp/total_i)
#print(accuracy)
np.save('BaselineTestVer_aug.npy', accuracy_list)
plt.plot(vertical_shift_range,accuracy_list)
plt.show()